Attempted Quality of Opponent Using Expected GF/GA There have been a lot of lists and rankings recently, Schwifty’s top 50 players, vBxnditt’s top 10 at each position, and killamilla6’s late nite stats (which did include quality of comp). A common theme is that it is very hard to accurately compare players in LG because of the varying level of competition. This was my *attempt* to compare some lines while using a method that considers the quality of competition. This method still has a lot of flaws, but I thought it was interesting to share anyway. The lines compared were all lines that hit 13 wins as of week 5, I am not saying these lines are the best, that was just my arbitrary cut-off. Expected GF/GA The idea is to calculate an expected GF (and GA) that would be scored if an average line were to play the matchups that each line played. Then compare that line’s true output to the expected output. This would be shown in GF above expected. The same can be calculated for GA above expected and GF-GA differential above expected. This would hopefully remove the huge stat boosts that are gained in blowout wins since teams constantly playing easy games will have a higher expected GF. It would also mean you won’t be faulted for having easy matchups. Even if you’re given easy matchups due to avail, by outscoring the expected GF you prove that you’re performing above average. Example: Team A scores 70 goals but had easy matchups with an expected GF of 65 Team B scores 50 goals, but with harder matchups with an expected GF of 40 Team A would rank lower despite having more goals, with a GF above of expected of 5 (70-65) compared to Team B’s 10 (50-40) The method to calculate the expected GF was to find the average GA of each opposing skater that the line had faced. This means if a line played 3 games, I would look at all 15 individual skaters they faced. For each opposing skater, I would take total goals against (by their team in games they played) and divide this by their GP. Then this expected GA per opposing player was averaged for every player that they played. Example: The opposing team is moose’s line, these are the stats used: The expected GA is calculated using each opposing player’s GF Moose: 62/15 = 4.13 Clutch: 62/15 = 4.13 Clutch: 62/15 = 4.13 Clutch: 62/15 = 4.13 PinPib: 24/6= 4.0 Average: (4.13+4.13+4.13+4.13+4) / 5 = 4.10 So, you would be expected to allow 4.10 GA facing this line, allowing less would put you above average. The calculation is the same for the expected GF. Results GF-GA above Expected / GP This shows how big of a goal differential a line had compared to expected. This should be most important as it accounts for both GF and GA. Higher differentials are better. GVNB’s 2.93 means they’re winning each game by about 3 more goals than an average team would with their schedule. GF above Expected / GP Actual GF – Expected GF Should show which teams are scoring more goals than expected, which lines are better purely on offence. GVNB’s 2.14 shows they’re scoring 2.14 more goals each game than an average team would with their schedule. GA above Expected / GP Actual GA – Expected GA Should show which teams are letting in fewer goals, essentially best defensive lines. Lower numbers are better. Skyre’s -1.89 means they’re letting in almost 2 fewer goals each game, compared to what an average team would let in with their schedule. Matchup Difficulty Expected GF-GA / GP Shows the expected goal differential of each team. Essentially shows which teams had the hardest and easiest matchups. Higher numbers indicate easier matchups. AB show’s 0.45 means that an average team facing their matchups would win each game by 0.45 goals on average. Rankings in Each All Stats Together The success of Expected GF/GA Balancing Sadly, it may have all been a waste of time. It is very hard to tell if comparing to GF above expected made a significant correction for quality comp, due to the small sample size. For the most part, the expected GF/GA did not cause much movement among lines compared to where they already ranked based on regular GF and GA stats. Expected GF helps differentiate some close ones, such as lowering AB Show’s line, and showing that there is a much larger gap between Rico and RJYorke’s line than shown from GF-GA. But for the top lines, it didn’t make much difference as none of these lines were obviously dodging. I think this stat could be useful if we had more lines included. There is more spread between GF/GA at the top shown here. I think using this for the middle of the pack teams, where GF/GA are much closer together, could be very useful to show which teams are being boosted by easy matchups and could help create a more balanced comparison between lines. It would make it easy to find cases of extreme dodging/scrub hunting. It’d also be interesting to use expected GA to see out of all lines which are the best defensively. This shows how the ranks of GF-GA and GF-GA above expected are almost the same (although small sample size) The following charts show how lines moved up or down rankings using expected stats compared to regular GF/GA. Green indicates they moved up a rank when changing to expected stats. Flaws Skill is so dependant on chem, linemates, and trades. Some players moving around so often causes for their GF/GA to be inaccurate of how good they are. GF/GA are not a perfect measure, you can have a low GF-GA if you are always playing good competition and winning 1-goal games. Opponents’ GF and GA are dependent on their matchups, high GF lines do not necessarily indicate better competition. Blowout games are still not accounted for perfectly. Winning 6-0 might hurt your GF if it’s a terrible line with expected 8ga. Some may argue you aren’t scoring as much as you should, some may argue it doesn’t matter how many if you win. Players’ stats were not available due to call-up, ban, or IDK why. This made it so some games had bad weightage. Ex. Hartford games had some lines with 3 players banned, so the expected GF/GA for those games are only weighted for 2 players rather than 5. Should have instead taken an average for each line then, averaged that to get the expected GF. Messed up and sometimes included ecus and sometimes didn’t. ‘Top lines’ are only judged by wins, not really the top lines. Would’ve liked to include more lines but don’t know how to auto calculate this. Lines compared For comparing each line, I picked one player that looked like they played most games with the line. Then I calculated the stats for only that player. Here is what (I think) each player's name should represent in terms of lines. Costello Line @I Costello l – SweatedOrb5 – MadhatterJokerr xFalsexHopex – Fleury I8I Rico505 Line Cross x I92I / IPastrnak 88I - @RiCo505 – Pearson x 70 Changing D Mooseballs04 Line @Mooseballs04 – aimNclu7ch – redsox162 Soxfan7744 – PinPib Skyre Line Architect3109 – @skyre5 – SouPerSaiyan BigD Auditor - McAvxy AB Show Line Cronkxy – AK11GKG – @The AB Show Gbaby Waffles - Elite x 18 GVNB Line @Gvnb – dooglare – Vrana x 15 xBeg I37I XI Tiger IX vBxnditt Line K6 I81I – IOTIS11I – AlexDaGreat78 Drover2? – @vBxnditt RJYorke Line @RJYorke – HabsAreReligion – Achilles Strife i wreckineyez I – mashle93 Mole11 Line Only here to show how an average line (6-5-1) compares to these lines. ------------------------------------------------------------------------------------------------------------------ Didn’t end up being as meaningful as I had hoped. Anyway, if anyone’s interested in seeing how their line compares in these stats, I’d be happy to add them when I have time. Thanks to @Fleury l8l for the proofread.
Great job orb and great read ! All media is good some Arnt gonna like or agree but thahs what makes media great !
appreciate the media aswell as the stats breakdown for me and others. If possible, would you be able to take out my 0-3 week and calculate my stats like that and compare them to other lines? pretty sure nearly half of my GA were in that week and id love to see what my average is like with that week taken out. If not all good, great post I love seeing the numbers broken down like this. looking forward to more in the future
@SweatedOrb5 great stuff pal, didn't know you were this deep into analytics. Was an interesting read, glad to see the homies @Mooseballs04 @redsox162 and @PinPib crushing it!
It’s an interesting read orbie, this is kinda setting what a median is with goals for and against and dividing deeper, but there could be things that scew it, just like when I got lucky and got 9 points that’s going to increase the GF but still not determine my opponent ability of an average 2 point game
I know how much time you must have poured into this and I'm getting nightmares even thinking about how this could be done for every player in the league at the end of the season so you could achieve that larger sample size you mentioned.....this is cool stuff tho,...i love this nerdy shit! If only we could pull stats off this site into a spreadsheet that would recognize the stats, oppsing team, opposing player + their stats....it'd be great once the formulas and template were in place....but yeah i like this #'s media!
agreed, definitely wish there was a better way to access the stats. wanted to do smth like this for a while but thought it'd be too much work. then when you said you went player by player to find your quality of competition I thought maybe it was possible. appreciate the work put into your show even more than before after seeing the time it takes
Just added it as a separate line as GVNB (13-0) so you can see them in the same table together. Best line (of these 9) by a longshot by these stats.
Should take a look at that Belleville scrubhunt line, would be interesting to see both sides of the spectrum
what line is this? this stat has a lot of problems, id like to try some different stuff and would want to test on a scrubhunt line